Material and methods

Study site
The study was carried out during the years 2016 and 2017 in an even-aged, pure beech stand (Fagus sylvatica L.) located at Selva Piana (Collelongo, Abruzzi Region, Italy 41°50’58” N, 13°35’17” E, 1560 m elevation) included in a 3000 ha forest within the external belt of Abruzzo-Lazio-Molise National Park (Central Apennine). The last dendrometric survey (2017) assessed a stand density of 725 trees ha−1, a basal area of 45.77 m2ha−1, a mean diameter at breast height (DBH) of 28.5 cm and a mean tree height of 23 m. In 2013, mean tree age was estimated to be about 110 years. The soil is humic alisol with a variable depth (40–100 cm), developed on calcareous bedrock. For the period 1989–2014, the mean annual temperature was 7.2°C, and the mean annual precipitation was 1178 mm, of which 10% concentrated during the summer (Guidolotti, Rey, D’Andrea, Matteucci & De Angelis 2013; Collaltiet al. 2016; Rezaie et al. 2018; Reyer et al.2020). The experimental area is part of the LTER network (Long Term Ecological Research) since 1996.
Climate and Phenology
The temperature and precipitation for the period 1989-2015, available on the Fluxnet2015 release, were used to characterize the, on average, climate conditions of the site. For the data gaps occurred during the experimental trial (2016-2017), we used the ERA5 database produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) (https://www.ecmwf.int/en/forecasts/datasets/archive-datasets/reanalysis-datasets/era5, data accessed: [12/04/2018]), according to the Fluxnet2015 release formulations (Pastorello et al. 2020). To evaluate peculiarities of the climatic conditions in 2016 and 2017 we calculated monthly differences with respect to the average values of precipitation and temperature observed in the site in the historical time series 1989-2015.
Leaf phenology was monitored using the MODIS Leaf Area Index product (LAI, MOD15A2H product, https://modis.gsfc.nasa.gov/ ) with 8-days temporal resolution and 500-meters spatial resolution (Myneni et al. , 2015). Critical dates, representing approximately linear transitions from one phenological phase to another, were identified and defined according to Zhang et al. (2003) as: (1) green-up , photosynthetic activity onset; (2) maximum LAI , supposed to be the leaf maturity phase; (3) senescence , sharp decrease of photosynthetic activity and green leaf area; (4) winter dormancy . In 2016, the leafless period after the late frost was identified from the day of the extreme event and the second green up.
Selection, measurements and sampling of trees
Five trees were selected according to their similarity with site tree ring chronology, the trees had a DBH ranging from 49 to 53 cm and a mean age of 109 ± 4 years. Trees were monitored from April 2016 to November 2017. Intra-annual radial growth of each selected tree was measured using permanent girth bands with 0.1 mm accuracy (D1 Permanent Tree Girth, UMS, Germany). Furthermore, stem diameter was recorded at the moment of each sampling of xylem for biochemical analyses (20 sampling dates from April 2016 to November 2017).
From each tree, micro-cores (2 mm diameter, 15 mm long) of wood were collected after bark removal, using the Trephor tool (Rossi, Menardi, Fontanella & Anfodillo 2005). All samples for biochemical analyses were immediately placed in dry ice for transport to the laboratory, then stored at −20 °C and, finally, stabilized through lyophilisation processes until NSCs analysis.
Daily radial increment (Ri, μm day–1), was calculated as follow:
\(R_{i}=\frac{R_{t}-R_{t-1}}{t}\) eq.1
where R is the radius of each i tree (μm), t is the date of sampling, and Δt is the time interval between the two sampling dates expressed in days.
In November 2017, at the end of the experimental trial, increment cores were collected at breast height from each tree, using an increment borer. Tree ring width series were converted into tree basal area increment (BAI, cm2 year–1), according to the following standard formula:
\(BAI=\pi\ \left(R_{n}^{2}-R_{n-1}^{2}\right)\) eq.2
with n being the year of tree-ring formation.
Starch and soluble sugar concentrations analysis
The freeze-dried xylem samples were milled to a fine powder and used for all analytical tests. For analysis of glucose, fructose, sucrose and starch, 10 mg of dry xylem powder were extracted in 1 ml of 80% ethanol/water at 80 °C for 45 minutes. After centrifugation at 16,000 x g for 5 minutes, soluble sugars were recovered in the supernatant while the pellet was resuspended in 1 ml of 40 mM acetate buffer (pH 4.5), then re-centrifuged 16,000 x g for 5 minutes. This procedure was repeated 4-times. The final pellet was autoclaved for 45 minutes at 120 °C in the same wash buffer. Enzymatic starch hydrolysis and the following glucose spectrophotometric assay were done as described by Moscatello et al. (2017). The supernatant solution containing soluble sugars was filtered on 0.2 μm nylon filters (GE-Whatman, Maidstone, UK), then analyzed by high-performance anion exchange chromatography with pulsed amperometric detection (HPAEC-PAD) (Thermo Scientific™ Dionex™ ICS-5000, Sunnyvale, CA U.S.A.)(Proietti et al. 2017).
Modelling of Intra-annual dynamics of non-structural carbohydrates
To evaluate the effects of the spring late frost (2016) and the heat wave and drought stress (2017) on the intra-annual NSCs dynamic, a representative benchmark of the typical intra-annual carbohydrates dynamic of the study site was needed. With this aim, a dataset on NSCs dynamic derived from other experimental trials at the site was used (Supporting Information Table S1). Dataset was composed of data of different years (i.e.: 2001, 2002, 2013, 2014, 2015, and 2018). This dataset included 39 observations of starch dynamic and 28 observations for both soluble sugars (glucose, fructose and sucrose) and total NSCs dynamic. Observations for soluble sugars were lower, because of the methodological sampling procedure used in 2015. During that year, woody samples were collected for xylogenesis analysis and maintained in ethanol-formalin acetic acid solution (FAA). Unfortunately, this methodology caused the loss of soluble sugars, while the starch integrity was preserved, as verified by means of specific analytical tests on woody tissues.
Different models based on data of starch, soluble sugars and total NSCs were used looking for possible patterns within the years and tested through the Akaike Information Criterion (AIC) (Akaike 1974; Aho, Derryberry & Peterson 2014) to select the simplest model able in reproducing the in situ observed pattern. The AIC quantifies the trade-off between parsimony and goodness-of-fit in a simple and transparent manner, estimating the relative amount of information lost by a given model. Hence, the model showing the lowest AIC is considered the model with the smallest information loss and, potentially, the most representative one (Akaike 1974). The four assumptions of linear model (homoscedasticity, normality of the error distribution, statistical independence of the errors and absence of influential points) were tested graphically (Fig. S1 - 3).
Statistical data analysis
Intra annual differences among contents of starch and total sugars were tested using one-way repeated measures analysis of variance (one factor repetition), using sampling date as predictive factor. The measured data of soluble sugars did not pass the normality test and were analysed by Repeated Measures Analysis of Variance on Ranks. Multiple comparisons were performed by the Student-Newman-Keuls Method.
Linear mixed models, considering “tree” and “sampling date” as crossed random effects, were used to account for the random variation of inter-annual starch, soluble and total sugar contents. Statistical analysis and figures were made using R 3.5.0 (R Development Core Team 2018).
Differences among modelled and measured data were identified using the interval of confidence (1.96 Standard error, SE), the lack of overlap between the two intervals of confidence indicates likely a statistically significant difference at the 95% level (P-value<0.05).